Getting Started
Prerequisites
- Node.js 24 or newer, which includes npm
- Access to an AI model: a ChatGPT subscription, a Vercel account, or an AI Gateway, OpenAI, or Anthropic API key. You don't need a Vercel project to start chatting.
Choose a model, provider, and channel that meet your data-processing and compliance requirements.
Create a project
Run eve init with a project name to create a project with an agent in agent/:
The command creates the project, installs dependencies, initializes Git, and opens the terminal UI.
From the root of a project that has a package.json and no agent/ directory, run:
The command adds an agent in agent/, installs dependencies, and opens the terminal UI.
eve connects with credentials you already have, or opens /login to set one up.
Once connected, you can start chatting with your agent. Use /model to switch models or reasoning effort.
Run the agent
The terminal UI opens after interactive initialization. To return later, run this from the project root:
eve reloads your changes as you work.
Organize your project
See Project Structure when adding another root agent or deciding where a frontend belongs. The Agent Files reference lists supported files and discovery rules.
Install manually
If you do not want to use the scaffold, install the runtime dependencies:
Declare Node.js 24 in package.json, then create agent/instructions.md and, when you need runtime configuration, agent/agent.ts.
The installed package's node_modules/eve/CHANGELOG.md contains that version's release notes and a link to the full changelog through that version. For an upgrade across multiple releases, review each version between the installed and target versions in the target version's changelog at https://raw.githubusercontent.com/vercel/eve/refs/tags/eve%40<version>/packages/eve/CHANGELOG.md.
Next steps
Try these from the eve terminal UI:
- Change your agent's behavior. Edit
agent/instructions.md, or ask the agent to update its own instructions. See Self-Modification. - Give your agent a new capability. Ask the agent to build one, such as "Add a tool that looks up the weather for a city" or "Connect to Linear."
- Let your agent communicate on another channel. Run
/addand choose a platform such as Slack or iMessage. - Deploy it. Run
/deployto ship the agent to Vercel, or see Self-Hosting.
For examples of what to build, see Use Cases. Then continue with the task you need:
| Goal | Read |
|---|---|
| Give the model some code it can run | Tools |
| Teach the agent a reusable procedure | Skills |
| Require human approval before a tool runs | Human-in-the-Loop |
| Connect the agent to an external MCP or OpenAPI service | Connections |
| Communicate with users through Slack, Discord, or another platform | Channels |
| Build a browser interface | Frontend Frameworks |
| Test agent behavior | Evals |
| Secure and deploy the agent | Authentication, then Deployment |
Read Execution Model and Durability to learn how sessions, turns, and steps run, and how waiting work survives restarts.
